A minimalistic program implementing Koza-style (tree-based) genetic programming to solve a symbolic regression problem.
tiny-gp.py is a basic (and fully functional) version, which produces textual output of the evolutionary progression and evolved trees.
tiny-gp-plus.py displays dynamic graphs of error and mean tree size (size = number of nodes), has a bloat-control option, and produces nicer, graphic output (you'll need to install https://pypi.org/project/graphviz/).
| Symbolic | Regression using GP |
|---|---|
| Objective | Find an expression with one input (independent variable x), whose output equals the value of the quartic function x4 + x3 + x2 + x + 1 |
| Function set | add, sub, mul |
| Terminal set | x, -2, -1, 0, 1, 2 |
| Fitness | Inverse mean absolute error over a dataset of 101 target values, normalized to [0,1] |
| Paremeters | POP_SIZE (population size), MIN_DEPTH (minimal initial random tree depth), MAX_DEPTH (maximal initial random tree depth), GENERATIONS (maximal number of generations), TOURNAMENT_SIZE (size of tournament for tournament selection), XO_RATE (crossover rate), PROB_MUTATION (per-node mutation probability) |
| Termination | Maximal number of generations reached or an individual with fitness = 1.0 found |
| Evolved solution | Another evolved solution |
|---|---|
 |
 |
| Bloat control | No bloat control |
|---|---|
 |
 |
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent